Diffusion-weighted MRI (DWI) is essential for stroke diagnosis, treatment decisions, and prognosis. However, image and disease variability hinder the development of generalizable AI algorithms with clinical value. We ...
详细信息
Diffusion-weighted MRI (DWI) is essential for stroke diagnosis, treatment decisions, and prognosis. However, image and disease variability hinder the development of generalizable AI algorithms with clinical value. We address this gap by presenting a novel ensemble algorithm derived from the 2022 Ischemic Stroke Lesion Segmentation (ISLES) challenge. ISLES’22 provided 400 patient scans with ischemic stroke from various medical centers, facilitating the development of a wide range of cutting-edge segmentation algorithms by the research community. By assessing them against a hidden test set, we identified strengths, weaknesses, and potential biases. Through collaboration with leading teams, we combined top-performing algorithms into an ensemble model that overcomes the limitations of individual solutions. Our ensemble model combines the individual algorithms’ strengths and achieved superior ischemic lesion detection and segmentation accuracy (median Dice score: 0.82, median lesion-wise F1 score: 0.86) on our internal test set compared to individual algorithms. This accuracy generalized well across diverse image and disease variables. Furthermore, the model excelled in extracting clinical biomarkers like lesion types and affected vascular territories. Notably, in a Turing-like test, neuroradiologists consistently preferred the algorithm’s segmentations over manual expert efforts, highlighting increased comprehensiveness and precision. Validation using a real-world external dataset (N=1686) confirmed the model’s generalizability (median Dice score: 0.82, median lesion-wise F1 score: 0.86). The algorithm’s outputs also demonstrated strong correlations with clinical scores (admission NIHSS and 90-day mRS) on par with or exceeding expert-derived results, underlining its clinical relevance. This study offers two key findings. First, we present an ensemble algorithm that detects and segments ischemic stroke lesions on DWI across diverse scenarios on par with expert (neuro)rad
Aims. We present a campaign designed to train the GRANDMA network and its infrastructure to follow up on transient alerts and detect their early afterglows. In preparation for O4 II campaign, we focused on GRB alerts ...
详细信息
Aims. We present a campaign designed to train the GRANDMA network and its infrastructure to follow up on transient alerts and detect their early afterglows. In preparation for O4 II campaign, we focused on GRB alerts as they are expected to be an electromagnetic counterpart of gravitational-wave events. Our goal was to improve our response to the alerts and start prompt observations as soon as possible to better prepare the GRANDMA network for the fourth observational run of LIGO-Virgo-Kagra (which started at the end of May 2023), and future missions such as SM. Methods. To receive, manage and send out observational plans to our partner telescopes we set up dedicated infrastructure and a rota of follow-up adcates were organized to guarantee round-the-clock assistance to our telescope teams. To ensure a great number of observations, we focused on Swift GRBs whose localization errors were generally smaller than the GRANDMA telescopes' field of view. This allowed us to bypass the transient identification process and focus on the reaction time and efficiency of the network. Results. During'Ready for O4 II', 11 Swift/INTEGRAL GRB triggers were selected, nine fields had been observed, and three afterglows were detected (GRB 220403B, GRB 220427A, GRB 220514A), with 17 GRANDMA telescopes and 17 amateur astronomers from the citizen science project Kilonova-Catcher. Here we highlight the GRB 220427A analysis where our long-term follow-up of the host galaxy allowed us to obtain a photometric redshift of z = 0.82 ± 0.09, its lightcurve elution, fit the decay slope of the afterglows, and study the properties of the host galaxy. Conclusions. During this 8-week-long GRB follow-up campaign, we successfully fulfilled our goal of training telescope teams for O4 and the improvement of the associated technical toolkits. For seven of the GRB alerts, our network was able to start the first observations less than one hour after the GRB trigger time. We also characterized the network effic
暂无评论